Hello,

I am trying to analyze methylationdata from TCGA Illumina 27k methylation arrays. I am using read.metharray from minfi to import the data and it creates a RGChannesSet:

> rgSet27
class: RGChannelSet 
dim: 55300 295 
metadata(0):
assays(2): Green Red
rownames(55300): 10008 10010 ... 7650762 7650767
rowData names(0):
colnames(295): 4321207002_A 4321207002_B ... 5543207071_J 5543207071_L
colData names(0):
Annotation
  array: IlluminaHumanMethylation27k
  annotation: ilmn12.hg19

From this I can calculate Beta values using getBeta but when I try to get the detection P values using the detectionP function I get "NA" values only. I think the problem lies somewhere in the Addresses for the negative controls which are required for the dectionP values to my understanding. For the Illumina 450k array everything works fine. Calling getControlAddress(rgSet, controlType = "NEGATIVE") gets me a number of addresses with a RGChannelSet from a 450k array while it just gets me

> getControlAddress(rgSet27, controlType = "NEGATIVE")
character(0)

for the RGChannelSet from the 27k arrays. This function is internally called by the detectionP function so I suspect somewhere there to be the problem.

> getControlAddress(IlluminaHumanMethylation27kmanifest,controlType = "NEGATIVE")
character(0)
> getControlAddress(IlluminaHumanMethylation450kmanifest,controlType = "NEGATIVE")
  [1] "13792480" "69649505" "34772371" "28715352" "74737439" "33730459" "54647418" "71684459" "40758350" "35678453" "33794466" "64607409"
 [13] "37639436" "44804473" "45735479" "74688409" "37748499" "15760434" "39647340" "46770480" "73794328" "13762418" "57645303" "44767440"
 [25] "74602472" "51736450" "61694431" "22623305" "34681355" "38610459" "71637302" "11634395" "52673469" "24725313" "62687391" "59623346".....

Thank you!

Unfortunately it's a problem with the case of the type names:

> zz <- as.list(IlluminaHumanMethylation27kmanifest@data)
> table(zz$TypeControl$Type)

          Background Bisulfite conversion            Extension 
                   8                   12                    4 
       Hybridization             Negative      Non-Polymorphic 
                   3                   32                    4 
 Normalization-Green    Normalization-Red             pACYC174 
                   8                    8                    6 
             phiX174                pUC19          Specificity 
                   6                    6                    8 
            Staining       Target Removal 
                   6                    1 

> zz4 <- as.list(IlluminaHumanMethylation450kmanifest@data)
> table(zz4$TypeControl$Type)

 BISULFITE CONVERSION I BISULFITE CONVERSION II               EXTENSION 
                     12                       4                       4 
          HYBRIDIZATION                NEGATIVE         NON-POLYMORPHIC 
                      3                     613                       4 
                 NORM_A                  NORM_C                  NORM_G 
                     32                      61                      32 
                 NORM_T             RESTORATION           SPECIFICITY I 
                     61                       1                      12 
         SPECIFICITY II                STAINING          TARGET REMOVAL 
                      3                       6                       2 

A cheap trick would be to run debug on detectionP, and when you get to this line:

controlIdx <- getControlAddress(rgSet, controlType = "NEGATIVE")

allow the line to run, and then overwrite with

controlIdx <- getControlAddress(rgSet, controlType = "Negative")

and then you should be able to proceed by just hitting the 'c' key.

This bug is fixed in minfi 1.29.4 (devel) and 1.28.4 (release) which I have just committed. Assuming they build fine, they will be available in a few days from Bioconductor and immediately from the GitHub page.

Thanks for the report and the diagnose.

Best, Kasper